English is Not All You Need: Systematically Exploring the Role of Multilinguality in LLM Post-Training

TL;DR

This paper systematically investigates how multilinguality during post-training affects large language models, showing that even minimal multilingual data benefits performance across languages and tasks.

Contribution

It provides a controlled study demonstrating the benefits of increased language coverage during post-training, especially for low-resource languages, and highlights the importance of multilinguality for cross-lingual transfer.

Findings

Increasing language coverage benefits all tasks and model scales.

Minimal multilinguality improves English performance and cross-lingual generalization.

Zero-shot transfer can match or surpass direct multilingual training effects.

Abstract

Despite the widespread multilingual deployment of large language models, post-training pipelines remain predominantly English-centric, contributing to performance disparities across languages. We present a systematic, controlled study of the interplay between training language coverage, model scale, and task domain, based on 220 supervised fine-tuning runs on parallel translated multilingual data mixtures spanning mathematical reasoning and API calling tasks, with models up to 8B parameters. We find that increasing language coverage during post-training is largely beneficial across tasks and model scales, with low-resource languages benefiting the most and high-resource languages plateauing rather than degrading. Even minimal multilinguality helps: incorporating a single non-English language improves both English performance and cross-lingual generalization, making English-only post-training largely suboptimal. Moreover, at sufficient language diversity, zero-shot cross-lingual transfer can match or exceed the effects of direct language inclusion in a low-diversity setting, although gains remain limited for typologically distant, low-resource languages.